Naver, which grew on search, shopping and payments, has gone all in on an artificial intelligence (AI) infrastructure business worth up to $10 billion (about 14.682 trillion won). The plan is for Nvidia to invest $1 billion (1.4682 trillion won) to secure 4.5% equity in Naver, while global alternative asset manager Brookfield supports financing of up to $9 billion (13.2174 trillion won).
Chair Lee Hae-jin of Naver's board aims to use this as a springboard to expand Naver from an internet platform into a global AI infrastructure corporations. But Naver must make the $10 billion financing a reality and secure long-term clients that will keep hundreds of megawatts (MW) of facilities running stably. At the same time, whether Naver can push ahead without a hitch with both its combination with Dunamu and large-scale AI investment is expected to determine the success or failure of the business.
◇ Nvidia's $1 billion is also "conditional"… $10 billion funding is the first gate
On the 27th, Naver said it will carry out a third-party paid-in capital increase of about $1 billion for Nvidia. Nvidia will acquire 7,241,564 new shares to secure 4.5% equity in Naver. The payment date is Oct. 30. This is Naver's first third-party paid-in capital increase in 22 years and the first since its transfer listing to KOSPI in 2008.
Nvidia's investment is seen less as a simple financial investment and more as a strategy to support Naver's expansion of AI infrastructure while securing long-term demand for hardware and software.
Naver has designated Brookfield as its exclusive capital partner and plans to expand the AI factory at the Sejong data center "Gak Sejong" from 55 MW in the first half of 2027 to 200 MW in 2028. Two hundred MW can accommodate about 100,000 Nvidia graphics processing units (GPUs). In the long term, Naver will build 1 gigawatt (GW)-class infrastructure to target sovereign AI demand from domestic and foreign corporations and governments.
Once the investment is completed, Nvidia will be both a major shareholder of Naver and a key partner supplying GPUs, network equipment and software to the AI factory. As the AI factory expands, not only will there be demand for new GPUs, but ongoing needs for equipment replacement and software use can also arise. For Nvidia, this offers the effect of securing a long-term sales channel while expecting an increase in equity value.
Similar aligned interests can be seen in its transaction with U.S. AI cloud firm CoreWeave. In January, Nvidia acquired an additional $2 billion worth of CoreWeave stock, and the two companies agreed to build more than 5 GW of AI factories by 2030. Nvidia signed a $6.3 billion contract to purchase cloud capacity that CoreWeave fails to sell to external clients. The structure reduces CoreWeave's utilization risk while expanding Nvidia-based computing supply.
In Naver's case, however, no demand backstop like CoreWeave's has been disclosed. There has been no confirmation that Nvidia will purchase Naver's unsold cloud capacity or guarantee purchases of a certain volume of GPUs. Above all, Nvidia's $1 billion investment is premised on Naver separately securing at least $9 billion in committed financing. Brookfield's support of up to $9 billion is also at a non-binding term sheet stage. The interest rate on the financing, collateral, repayment terms and risk-sharing arrangements will be decided in the definitive agreement. For now, it is closer to the start of large-scale financing talks than to having the full $10 billion secured.
◇ Securing clients to fill 200 MW of facilities is key… operating in-house differs from external sales
The profitability of an AI data center depends on utilization rather than the number of GPUs. Without long-term clients, massive capital expenditures on facilities, financing costs and electricity costs remain a burden.
Naver has operated its own GPU cluster, hyperscale data center, AI platform and in-house AI models. It also has experience handling large-scale internal demand in search, shopping, advertising and cloud. But the ability to operate its own services stably is different from the ability to sell AI infrastructure to external clients. It has yet to be proven whether existing domestic corporate clients alone can fill not only 200 MW but also 1 GW-class facilities. Naver must sign long-term contracts with global corporations or overseas government agencies to secure stable utilization.
There is also the burden of competing with operators such as Amazon Web Services (AWS), Microsoft and Google, which have data centers and large clients around the world. For Naver to be chosen as an alternative, it must demonstrate not only price competitiveness but also stable power supply, Data Sovereignty and security, overseas sales networks and a long-term operating track record. Naver CEO Choi Soo-yeon's remark that the company will "also quickly wrap up securing clients for the AI factory" suggests that work is underway to lock in key customers.
An IT industry source said, "For Naver to be chosen by global clients as a differentiated alternative to existing large cloud operators, it must prove not only price but also power security, service stability and the ability to respond to local regulations."
◇ Pushing AI factory and Dunamu combination in tandem… risk of dispersed execution capacity
Alongside AI infrastructure investment, Naver is also pushing to combine Naver Financial and Dunamu. The combination with Dunamu is a choice to expand a digital finance business spanning payments, finance and virtual assets. The burdens of the two businesses differ in nature. The AI factory is an infrastructure business that requires large-scale capital over a long period for data center construction, GPU procurement, power security and equipment replacement. By contrast, the Dunamu combination uses a stock-swap method and thus does not require the same level of capital expenditures. Instead, it comes with tasks such as reviews by financial authorities and the Korea Fair Trade Commission, responding to virtual asset regulations, and organizing governance and integrating organizations.
While the capital burden is concentrated on the AI factory, pushing ahead simultaneously with the Dunamu combination could disperse management's focus and the organization's execution capacity. If funding costs and depreciation for the AI data center grow more than expected while the Dunamu combination is delayed, Naver could shoulder different types of challenges all at once.
There are expectations that Dunamu's financial strength and customer base can be leveraged for the AI business. But whether fintech and virtual asset businesses can be direct demand sources for a 1 GW-class AI data center is another matter. No concrete plan has been presented for Dunamu to become a long-term client of the AI factory or for the two companies to jointly develop AI services.
For Naver to frame the combination with Dunamu as a synergy for the AI infrastructure business, it must show not only internal compute demand but also joint services, client acquisition and revenue models. Otherwise, pushing the two businesses simultaneously could be seen as a factor that increases execution burden rather than synergy.
Chair Lee Hae-jin's bet demands three answers: whether the conditionally presented $10 billion can actually be secured; whether long-term clients can be found to fill facilities expanding from 200 MW to 1 GW; and whether a management system can be put in place to push ahead with the combination with Dunamu and the AI infrastructure business at the same time.
Kim Kyung-won, a distinguished professor of business administration at Sejong University, said, "The names Nvidia and Brookfield are only a starting point," adding, "If Naver fails to prove actual fund deployment, long-term client contracts and high utilization, the $10 billion AI factory could remain a bet that saddles the company with massive expense rather than a new growth engine." He added, "A breakthrough will open only if Naver leverages the Nvidia brand to crack the Asian market."